Model benchmarks
Qwen3.6-27B-MLX-8bit local LLM performance
As of August 2026, Qwen3.6-27B-MLX-8bit runs at up to 15.3 tok/s for local inference (best of 9 community benchmark runs across 1 GPU).
oMLX
Model size
27B
Peak speed
15.3 tok/s
Average speed
10.6 tok/s
Min memory
32.8 GB
Max context
131,072 tokens
Avg runtime / run
18m 35s
Avg quality
81.1
Benchmark runs
9
GPUs tested
1
Quality by task
Average LLM-judged quality (0–100) with the run-to-run spread shown as a P5–P95 band, overall and for each benchmark task, across all 9 runs. The low and high columns show how much the judge’s score varies between runs, and need at least two runs to display.
| Task | P5 (low) | Avg | P95 (high) |
|---|---|---|---|
| Overall | 78.1 | 81.1 | 85.3 |
| Agent Workflow | 70.2 | 81.3 | 89.2 |
| Code Generation | 64.6 | 74.5 | 81.0 |
| Role Play & Narrative | 81.8 | 87.6 | 90.7 |
| Research & Analysis | 76.3 | 81.1 | 85.5 |
Performance by hardware and tool
Every hardware/tool/quantization combination Qwen3.6-27B-MLX-8bit has been benchmarked on, ranked by peak token generation speed. Last updated August 2026.
| Hardware | Tool | Quant | Best tok/s | Avg tok/s | Memory | Context | Quality | Runs |
|---|---|---|---|---|---|---|---|---|
| Apple M5 Max | oMLX | — | 15.3 tok/s | 10.6 tok/s | 32.8 GB | 131,072 tokens | 81.1 | 9 |
Frequently asked questions
- Is Qwen3.6-27B-MLX-8bit good for coding?
- In our benchmarks, Qwen3.6-27B-MLX-8bit scores 74.5/100 for coding. It runs at about 10.6 tok/s, so if you want more speed, Gemma4:E2B/QAT-MTP@131K is faster (~303.9 tok/s) and still scores well for coding (66.9/100).
- Is Qwen3.6-27B-MLX-8bit good for agentic (tool-using) tasks?
- In our benchmarks, Qwen3.6-27B-MLX-8bit scores 81.3/100 for agentic workflows. It runs at about 10.6 tok/s, so if you want more speed, mudler/Ornith-1.5-35B-A3B-APEX-MTP-GGUF is faster (~190.9 tok/s) and still scores well for agentic workflows (85.7/100).
- How fast is Qwen3.6-27B-MLX-8bit for local inference?
- Across 9 community benchmark runs, Qwen3.6-27B-MLX-8bit reaches up to 15.3 tok/s and averages 10.6 tok/s, with the fastest results on Apple M5 Max.
- How much memory does Qwen3.6-27B-MLX-8bit need?
- The leanest observed configuration used about 32.8 GB of memory.
- Which tools have been used to run Qwen3.6-27B-MLX-8bit?
- Benchmarks were submitted using oMLX. Results are community-contributed and updated as new runs arrive.